fix-ai-writing

name	fix-ai-writing
version	3.1.0
description	Use when editing prose, technical docs, or READMEs to remove AI writing patterns. Detects and rewrites 42 general patterns plus 7 technical-doc patterns. Based on Wikipedia's "Signs of AI writing" guide.
license	MIT
compatibility	any-agent
allowed-tools	["Read","Write","Edit","Grep","Glob","AskUserQuestion"]

Fix AI Writing

Identify and rewrite AI-generated text patterns. Based on Wikipedia:Signs of AI writing.

Process

1. Scan for the patterns below. For technical docs, also apply T1-T7.
2. Write a draft rewrite. Match the register. Vary sentence length. Prefer simple constructions (is/are/has) and concrete specifics.
3. Self-check: "What still reads as AI?" List remaining tells.
4. Revise into a final rewrite. Scan for — and – before returning (see §14).

Deliver: draft, "still AI" bullets, final rewrite, summary of changes.

Voice

If the user provides a writing sample, read it first. Note:

• Sentence length (short and punchy? long and flowing? mixed?)
• Word choice level (casual? academic?)
• How they start paragraphs and handle transitions
• Punctuation habits (parenthetical asides? semicolons?)
• Recurring phrases or verbal tics

Match their voice in the rewrite. If they write short sentences, don't produce long ones. If they use "stuff," don't upgrade to "elements."

To provide a sample: inline ("Here's a sample of my writing: [sample]") or by file path.

When no sample is provided and the content calls for it (blog posts, essays, opinion): have opinions, vary rhythm, let some mess in. For technical, legal, or reference text, neutral and plain is the correct voice.

Signs that clean text is still soulless:

• Every sentence is the same length and structure
• No opinions, just neutral reporting
• No acknowledgment of uncertainty or mixed feelings
• No first-person perspective when appropriate
• Reads like a Wikipedia article or press release

Soulless: "The experiment produced interesting results. The agents generated 3 million lines of code. Some developers were impressed while others were skeptical." Has a pulse: "I genuinely don't know how to feel about this one. 3 million lines of code, generated while the humans presumably slept. Half the dev community is losing their minds, half are explaining why it doesn't count."

CONTENT PATTERNS

1. Significance inflation

Watch for: stands/serves as, is a testament, vital/crucial/pivotal role, underscores importance, reflects broader, evolving landscape, indelible mark, deeply rooted

Problem: LLMs puff up importance with statements about how things represent or contribute to broader topics.

Before: "established in 1989, marking a pivotal moment in the evolution of regional statistics" After: "established in 1989 to collect and publish regional statistics independently"

2. Notability inflation

Watch for: independent coverage, national media outlets, active social media presence

Problem: Lists sources without context to hammer notability.

Before: "cited in The New York Times, BBC, Financial Times, and The Hindu" After: "In a 2024 New York Times interview, she argued that AI regulation should focus on outcomes"

3. Superficial -ing analyses

Watch for: highlighting, underscoring, ensuring, reflecting, symbolizing, contributing to, fostering, showcasing

Problem: Present participle phrases tacked onto sentences for fake depth.

Before: "resonates with the region's natural beauty, symbolizing Texas bluebonnets" After: "The architect said these colors reference local bluebonnets and the Gulf coast."

4. Promotional language

Watch for: boasts, vibrant, rich, profound, showcasing, nestled, in the heart of, groundbreaking, renowned, breathtaking, must-visit, stunning

Before: "Nestled within the breathtaking region, it stands as a vibrant town" After: "Alamata Raya Kobo is a town in the Gonder region, known for its weekly market"

5. Vague attributions

Watch for: Industry reports, Experts argue, Some critics argue, several sources

Problem: Opinions attributed to unnamed authorities.

Before: "Experts believe it plays a crucial role in the regional ecosystem" After: "supports several endemic fish species, according to a 2019 survey by the Chinese Academy of Sciences"

6. "Challenges and Future Prospects" formula

Watch for: Despite its... faces several challenges..., Despite these challenges, Future Outlook

Before: "Despite these challenges, Korattur continues to thrive as an integral part of Chennai's growth" After: "Traffic congestion increased after 2015 when three new IT parks opened"

LANGUAGE AND GRAMMAR PATTERNS

7. AI vocabulary words

High-frequency: additionally, align with, crucial, delve, enduring, enhance, fostering, garner, highlight (verb), interplay, intricate, key (adj), landscape (abstract), pivotal, showcase, tapestry (abstract), testament, underscore, valuable, vibrant

These words co-occur far more frequently in post-2023 text.

8. Copula avoidance

Watch for: serves as, stands as, marks, represents [a], boasts, features, offers [a]

Before: "Gallery 825 serves as LAAA's exhibition space... boasts over 3,000 square feet" After: "Gallery 825 is LAAA's exhibition space... has four rooms totaling 3,000 square feet"

9. Negative parallelisms

"Not only...but..." and "It's not just about..., it's..." are overused. Also tailing negation fragments: "no guessing," "no wasted motion."

Before: "It's not merely a song, it's a statement" After: "The heavy beat adds to the aggressive tone"

Before (tailing negation): "The options come from the selected item, no guessing." After: "The options come from the selected item without forcing the user to guess."

10. Rule of three

LLMs force ideas into groups of three.

Before: "keynote sessions, panel discussions, and networking opportunities" After: "The event includes talks and panels. There's also time for informal networking."

11. Synonym cycling

Repetition-penalty causes excessive synonym substitution.

Before: "The protagonist faces many challenges. The main character must overcome obstacles. The central figure eventually triumphs. The hero returns home." After: "The protagonist faces many challenges but eventually triumphs and returns home."

12. False ranges

"From X to Y" where X and Y aren't on a meaningful scale.

Before: "from the singularity of the Big Bang to the grand cosmic web" After: "covers the Big Bang, star formation, and current theories about dark matter"

13. Passive voice / subjectless fragments

Before: "No configuration file needed. The results are preserved automatically." After: "You do not need a configuration file. The system preserves the results automatically."

STYLE PATTERNS

14. Em dashes: cut them

Hard constraint. The final rewrite contains no em dashes (—) or en dashes (–). Replace with: period, comma, colon, parentheses, or restructure. Also catch spaced em dashes (—) and double hyphens (--).

Scan the final output for — and – before returning. Any hit means the draft isn't done.

15. Boldface overuse

Before: "blends OKRs, KPIs, and Business Model Canvas" After: "blends OKRs, KPIs, and visual strategy tools like the Business Model Canvas"

16. Inline-header vertical lists

Before:

• User Experience: The user experience has been significantly improved.
• Performance: Performance has been enhanced through optimized algorithms.
• Security: Security has been strengthened with end-to-end encryption.

After: "The update improves the interface, speeds up load times through optimized algorithms, and adds end-to-end encryption."

17. Title case in headings

Before: "## Strategic Negotiations And Global Partnerships" After: "## Strategic negotiations and global partnerships"

18. Emojis

Before: "🚀 Launch Phase: The product launches in Q3" After: "The product launches in Q3."

19. Curly quotation marks

Replace "..." with "...".

COMMUNICATION PATTERNS

20. Chatbot artifacts

Watch for: I hope this helps, Of course!, Certainly!, Would you like..., let me know

Before: "Here is an overview of the French Revolution. I hope this helps! Let me know if you'd like me to expand on any section." After: "The French Revolution began in 1789 when financial crisis and food shortages led to widespread unrest."

21. Knowledge-cutoff disclaimers / speculative gap-filling

Before: "While specific details about the company's founding are not extensively documented in readily available sources, it appears to have been established sometime in the 1990s." After: "The company was founded in 1994, according to its registration documents."

When a model can't find a source, it writes a paragraph about not finding one and invents plausible filler. Say what's unknown, or cut the sentence.

22. Sycophantic tone

Before: "Great question! You're absolutely right that this is a complex topic. That's an excellent point about the economic factors." After: "The economic factors you mentioned are relevant here."

FILLER AND HEDGING

23. Filler phrases

"In order to" → "To". "Due to the fact that" → "Because". "At this point in time" → "Now". "Has the ability to" → "Can". "It is important to note that" → cut.

24. Excessive hedging

Before: "It could potentially possibly be argued that the policy might have some effect" After: "The policy may affect outcomes"

25. Generic positive conclusions

"The future looks bright" / "Exciting times lie ahead" / "a major step in the right direction"

Replace with a concrete fact or cut.

26. Hyphenated word overuse

AI hyphenates uniformly, including in predicate position. Humans drop the hyphen after the noun.

Before: "The team is cross-functional, the report is high-quality, and the methodology is data-driven." After: "The team is cross functional, the report is high quality, and the methodology is data driven."

27. Persuasive authority tropes

"The real question is...", "At its core...", "What really matters is..."

Replace the formula with the concrete claim.

28. Signposting

"Let's dive in," "let's explore," "here's what you need to know"

Cut. Just say the thing.

29. Fragmented headers

A heading followed by a one-line paragraph restating the heading. Delete the restatement.

30. Diff-anchored writing

Writing that narrates a change ("This was added to replace...") instead of describing the current state. Unless the document is a changelog, write about what exists now.

31. Manufactured punchlines

Stacked short declarative fragments for drama. One short sentence for emphasis is fine. A run of them is engineered.

Before: "Then AlphaEvolve arrived. It had no preference for symmetry. No aesthetic prior. The old rules were gone." After: "AlphaEvolve did not favor symmetry or human-looking designs. That made some older assumptions less useful."

32. Aphorism formulas

"X is the Y of Z," "X becomes a trap," "the language of," "the architecture of"

Replace with the concrete claim.

33. Conversational rhetorical openers

"Honestly?", "Look,", "Here's the thing" as standalone hooks before an ordinary point. The tell is the theatrical pause-and-reveal, not the word itself.

34. Section summaries

A paragraph at the end of a section restating what it just said. Delete it.

35. Abrupt style shifts

Register, sentence length, or vocabulary changes sharply between adjacent sections. LLMs don't maintain voice across long outputs.

36-42. Quick checks

• 36. Markdown in non-Markdown contexts. **bold** in plain text, emails, or wiki markup.
• 37. Heading level skips. H2 to H4 without H3.
• 38. Horizontal rules before headings. --- before a section header. Redundant.
• 39. Placeholder language. [Insert X here], "as appropriate," "as applicable."
• 40. Fabricated references. DOIs, ISBNs, or URLs that don't resolve. Check if possible, flag if not.
• 41. Prompt refusal leaking. "I can't help with," "As an AI."
• 42. Abrupt cutoffs. Text stopping mid-sentence from token limits.

TECHNICAL DOC PATTERNS

Apply these when the target is a README, API doc, architecture doc, or runbook.

T1. Feature-list puffery

"Powerful," "robust," "seamless," "comprehensive." Describe what the thing does, not how great it is.

Before: "Our powerful CLI provides seamless integration with robust error handling" After: "The CLI reads your config, runs the migration, and logs each step to stdout"

T2. Fake simplicity

"Simply run..." / "Just add..." before a multi-step process.

Before: "Simply install the package and you're ready to go!" After: "Install the package. You'll also need a database connection (see Configuration)."

T3. Badge walls

15+ badges at the top of a README. Keep ones the reader needs to decide whether to use the project. Cut the rest.

T4. Aspirational architecture

Documenting the system you wish you had. Write what the code does today.

T5. Copy-paste API docs

Every endpoint: "This endpoint allows you to [verb] a [noun]. It accepts the following parameters..." Vary the structure or just show request/response.

T6. Changelog theater

"Improved performance," "enhanced stability," "fixed minor bugs." If you can't name the change, the entry is filler.

Before: "Improved overall performance" After: "Reduced cold-start time from 4s to 800ms by lazy-loading the config parser"

T7. Redundant overview sections

Before: "## Overview\nMyTool is a command-line tool for managing database migrations. It provides..." After: "MyTool manages database migrations from the command line."

FALSE POSITIVES

Single instances of any pattern are not reliable tells. Look for clusters. Do not flag:

• Perfect grammar (professionals exist)
• Formal vocabulary (§7 targets specific AI words, not all fancy words)
• Em dashes alone (common among editors; flag only with other tells)
• One short emphatic sentence (flag only stacked runs)
• Curly quotes alone (OS auto-curl)
• Unsourced claims (most of the web is unsourced)
• Secondhand text (don't rewrite watched phrases inside quotations, titles, or examples)

A single em dash means nothing. Em dashes plus rule-of-three plus vibrant tapestry plus a "Conclusion" section is a confession.

Preserve signs of human writing: specific hard-to-fabricate details, mixed feelings, era-bound references, genuine asides and self-corrections, varied sentence length.

Example

See example-lisbon.md in this directory for a full before/draft/audit/final walkthrough.

Reference

Based on Wikipedia:Signs of AI writing, maintained by WikiProject AI Cleanup. Patterns 34-42 and T1-T7 extend the Wikipedia source.